How does a country's economic status impact it's football performance in the international stage?

By Nitan Singh, Vladamir Rife, Ehaab Basil

SETUP AND DATA WRANGLING

Football (soccer) is the most popular sport on Earth, and as such, many countries at many different economic levels compete to be the best. However, some countries are obviously better than others in the sport. This project aims to see if the economy and stability of a country correlates signifigantly with the performance of their national team.

Our hypothesis is that countries with a greater economic performance tend to have better national teams than those that do not. In order to test this, we will use machine learning to analyze each nation's GDP (Gross Domestic Product), GDP-per-captia, and PPP (Purchasing Power Parity), compared to their total FIFA points as the dependent variable. We will only be analyzing this for the year of 2016, as that is the latest year across all of the datasets.

Setup

Below are the packages and libraries we will use for working with our data.

import pandas as pd
import numpy as np
from sklearn import linear_model
from datetime import datetime
import matplotlib.pyplot as plt
import statsmodels.api as sm
import seaborn as sns
pd.reset_option('^display.', silent=True)

Data Collection and Curation

The datasets we will use for this project include one table with each country's GDP, a second table with each country's population, a third table with each country's PPP, and a fourth table with each country's fragility index all acting as independent varaibles. and a final table holding each country's FIFA rankings and total points as our dependent variable. The links to each data source are provided below:

GDP: https://github.com/datasets/gdp/blob/master/data/gdp.csv

Population: https://github.com/datasets/population/blob/master/data/population.csv

PPP: https://github.com/datasets/ppp/blob/master/data/ppp-gdp.csv

FIFA ranking: https://github.com/cnc8/fifa-world-ranking/blob/master/fifa_ranking-2020-12-10.csv

Why FIFA Total Points?

We chose the total points listed in the FIFA rankings database to be our measure of each country's performance in football since it is the most widely used metric for ranking national teams in men's soccer. The basic algorithm for computing the total points that produce these rankings is as follows:

The country's total points = (Their previous 12 month averaged match points + Their 36 month averaged match points before that)

Furthermore, the point system for each match is as follows:

Points from a single match = M x I x T x C

M = 3, 1, 0, 1, or 2 depending on if the country won, drew, lost, lost on penalties, or won on penalties respectively

I = 1, 2.5, 3, or 4 depending on if the match was a friendly, world-cup/confederation-level qualifier, confederation-level final, or world cup final game respectively

T = (200 - the opposing team's FIFA rank)

C = 1, 0.99, or 0.85 depending on if the opposing team is in the CONMEBOL, UEFA, or AFC/CAF/OFC/CONCACAF confederations respectively

More information about this algorithm exists at: https://www.goal.com/en-us/news/fifa-world-ranking-how-it-is-calculated-what-it-is-used-for/16w60sntgv7x61a6q08b7ooi0p

Why GDP?

We chose GDP (Gross Domestic Product) as a metric for a country's wealth since it is a direct measurement of a country's monetary productivity and consumption. The basic algorithm for computing GDP is as follows:

The country's GDP = C + G + I + NX

C = total domestic consumer spending or private consumption expenditure

G = total government consumption expenditure and gross investment carried out under the government’s name

I = all capital expenditures or private domestic investments

NX = (country's total export costs - country's total import costs)

More information about this algorithm exists at: https://www.analyticssteps.com/blogs/introduction-gross-domestic-product-gdp

Why GDP Per Capita?

We chose GDP per capita as a metric for a country's wealth since it is a measurement of a country's monetary productivity and consumption as opposed to its population. The basic algorithm for computing GDP per capita is as follows:

The country's GDP Per Capita = (The country's GDP / The country's population)

Why PPP?

PPP (Purchasing Power Parity) is the measure of how accessible and affordable goods are in a country with one standard unit of their currency (Ex: USD, Euro, Russian Ruble, etc). We chose this as a metric for a country's wealth since it is a good measure of how financially comfortable it is for citizens to exist in their country of origin. PPP can be calculated using the following formula:

The country's PPP = Cost of goods in the country / Cost of goods in another country (usually the US)

More information about this algorithm exists at: https://www.ig.com/en/trading-strategies/what-is-purchasing-power-parity--ppp---191106

Read the data from each of the github tables into a dataframe. We are only interested in the rows concerning individual countries, so we will delete any row with information about a country grouping (Ex: "Arab World"). We also need to drop any unecessary columns or rename any columns that might conflict with the other tables we will merge that table with.

#Read in the country GDP data from the github database, group them by year,
#drop all the irrelevant rows and columns, replace the Value label with GDP, and
#print the resulting gdp_df head
gdp_df = pd.read_csv('https://github.com/datasets/gdp/raw/master/data/gdp.csv')
gdp_df.groupby('Year')
gdp_df = gdp_df.drop(labels=range(0, 2306), axis=0)
gdp_df = gdp_df.drop(columns='Country Code')
gdp_df.rename(columns={'Value':'GDP'}, inplace=True)
gdp_df.head()

	Country Name	Year	GDP
2306	Afghanistan	1960	5.377778e+08
2307	Afghanistan	1961	5.488889e+08
2308	Afghanistan	1962	5.466667e+08
2309	Afghanistan	1963	7.511112e+08
2310	Afghanistan	1964	8.000000e+08

#Read in the country population data from the github database, group them by year,

population_df = pd.read_csv('https://github.com/datasets/population/raw/master/data/population.csv')
population_df.groupby('Year')
#drop all the irrelevant rows and columns, replace the Value label with population,
#and print the resulting population_df head
population_df = population_df.drop(labels=range(0, 2714), axis=0)
population_df = population_df.drop(columns='Country Code')
population_df.rename(columns={'Value':'Population'}, inplace=True)
population_df.head()

	Country Name	Year	Population
2714	Afghanistan	1960	8996973
2715	Afghanistan	1961	9169410
2716	Afghanistan	1962	9351441
2717	Afghanistan	1963	9543205
2718	Afghanistan	1964	9744781

#Read in the country PPP data from the github database, group them by year,

ppp_df = pd.read_csv('https://github.com/datasets/ppp/raw/master/data/ppp-gdp.csv')
#drop all the irrelevant columns, replace the Value label with population,
#and print the resulting ppp_df head
ppp_df = ppp_df.drop(columns='Country ID')
ppp_df.rename(columns={'Country':'Country Name'}, inplace=True)
ppp_df.head()

	Country Name	Year	PPP
0	Kenya	2019	40.975445
1	Kenya	2018	40.179628
2	Kenya	2017	40.184937
3	Kenya	2016	39.384956
4	Kenya	2015	39.040478

#Read in the FIFA ranking data from the github database, sort it by the country
#name, make a Year column including only the year from rank_date, rename the 
#column country_full to be Country Name
FIFA_df = pd.read_csv(r"https://raw.githubusercontent.com/cnc8/fifa-world-ranking/master/fifa_ranking-2020-12-10.csv")
FIFA_df.sort_values("country_full")
FIFA_df['Year'] = FIFA_df['rank_date'].apply(lambda x: x[:4])
FIFA_df = FIFA_df.rename(columns={"country_full": "Country Name"})
#Sort the dataframe by year, drop the rank_date column, and accomidate the table
#to be able to hold the US
FIFA_df.sort_values('Year')
FIFA_df = FIFA_df.drop(columns='rank_date')
FIFA_df.loc[(FIFA_df['Country Name'] == 'USA'),'Country Name']='United States'

#Some countries have different names on different databases, so we should rename them so they have 
#consistent names, since we will use the country names to merge later on 
FIFA_df.loc[(FIFA_df['Country Name'] == 'USA'),'Country Name']='United States'

gdp_df.loc[(gdp_df['Country Name'] == 'Congo, Dem. Rep.'),'Country Name']='Congo DR'
gdp_df.loc[(gdp_df['Country Name'] == 'Congo, Rep.'),'Country Name']='Congo'

ppp_df.loc[(ppp_df['Country Name'] == 'Congo, Dem. Rep.'),'Country Name']='Congo DR'
ppp_df.loc[(ppp_df['Country Name'] == 'Congo, Rep.'),'Country Name']='Congo'

population_df.loc[(population_df['Country Name'] == 'Congo, Dem. Rep.'),'Country Name']='Congo DR'
population_df.loc[(population_df['Country Name'] == 'Congo, Rep.'),'Country Name']='Congo'

#Set proper types for year columns
FIFA_df['Year'] = FIFA_df['Year'].astype(int)
gdp_df['Year'] = gdp_df['Year'].astype(int)

#merge FIFA dataframe and GDP dataframe into one dataframe, called master_df. This will be the main dataframe that we reference
#for the remainder of the tutorial
#We will always do an inner merge, so we dont have any rows that dont have certain column values
master_df = FIFA_df.merge(gdp_df, how='inner', on=['Country Name', 'Year'])
master_df = master_df.sort_values(['Country Name', 'Year'])
master_df.head()

	id	rank	Country Name	country_abrv	total_points	previous_points	rank_change	confederation	Year	GDP
17586	20000	204	Afghanistan	AFG	7	0	0	AFC	2003	4.583644e+09
17587	20000	203	Afghanistan	AFG	9	7	1	AFC	2003	4.583644e+09
17588	20000	198	Afghanistan	AFG	48	9	5	AFC	2003	4.583644e+09
17589	20000	198	Afghanistan	AFG	48	48	0	AFC	2003	4.583644e+09
17590	20000	199	Afghanistan	AFG	48	48	1	AFC	2003	4.583644e+09

#merge population dataframe into master_df
master_df = master_df.merge(population_df, how='inner', on=['Country Name', 'Year'])
master_df.head()

	id	rank	Country Name	country_abrv	total_points	previous_points	rank_change	confederation	Year	GDP	Population
0	20000	204	Afghanistan	AFG	7	0	0	AFC	2003	4.583644e+09	23680871
1	20000	203	Afghanistan	AFG	9	7	1	AFC	2003	4.583644e+09	23680871
2	20000	198	Afghanistan	AFG	48	9	5	AFC	2003	4.583644e+09	23680871
3	20000	198	Afghanistan	AFG	48	48	0	AFC	2003	4.583644e+09	23680871
4	20000	199	Afghanistan	AFG	48	48	1	AFC	2003	4.583644e+09	23680871

#merge PPP dataframe into master_df

master_df = master_df.merge(ppp_df, how='inner', on=['Country Name', 'Year'])
master_df

	id	rank	Country Name	country_abrv	total_points	previous_points	rank_change	confederation	Year	GDP	Population	PPP
0	20000	204	Afghanistan	AFG	7	0	0	AFC	2003	4.583644e+09	23680871	10.074357
1	20000	203	Afghanistan	AFG	9	7	1	AFC	2003	4.583644e+09	23680871	10.074357
2	20000	198	Afghanistan	AFG	48	9	5	AFC	2003	4.583644e+09	23680871	10.074357
3	20000	198	Afghanistan	AFG	48	48	0	AFC	2003	4.583644e+09	23680871	10.074357
4	20000	199	Afghanistan	AFG	48	48	1	AFC	2003	4.583644e+09	23680871	10.074357
...	...	...	...	...	...	...	...	...	...	...	...	...
41848	43892	110	Zimbabwe	ZIM	338	332	2	CAF	2016	1.661996e+10	14030390	0.521862
41849	43892	113	Zimbabwe	ZIM	316	338	3	CAF	2016	1.661996e+10	14030390	0.521862
41850	43892	110	Zimbabwe	ZIM	323	316	3	CAF	2016	1.661996e+10	14030390	0.521862
41851	43892	102	Zimbabwe	ZIM	334	323	8	CAF	2016	1.661996e+10	14030390	0.521862
41852	43892	102	Zimbabwe	ZIM	334	334	0	CAF	2016	1.661996e+10	14030390	0.521862

41853 rows × 12 columns

#Drop any duplicates in country and year, to trim the dataset and make it faster to work on
master_df.drop_duplicates(subset=['Country Name', 'Year'], keep='first', inplace=True)
master_df = master_df.sort_values(['Country Name', 'Year'])
#country_df = master_df.loc[master_df['Country Name'].str.contains('United', case=False)]
#country_df

#Calculate GDP per capita as GDP/Population, create a new column that stores this value
master_df['GDP_capita'] = master_df['GDP'] / master_df['Population']

#Isolate dataframe to only year of 2016, which is the latest year in the dataset
master_df = master_df.loc[master_df['Year']==2016]
temp  = master_df.loc[master_df['Year']==2016]

master_df

	id	rank	Country Name	country_abrv	total_points	previous_points	rank_change	confederation	Year	GDP	Population	PPP	GDP_capita
154	20000	153	Afghanistan	AFG	183	194	3	AFC	2016	1.946902e+10	35383128	17.445828	550.234626
423	43932	38	Albania	ALB	746	746	0	UEFA	2016	1.186387e+10	2876101	42.385879	4124.982390
692	43843	28	Algeria	ALG	845	845	0	CAF	2016	1.590491e+11	40551404	37.155827	3922.160050
961	43844	105	Angola	ANG	333	330	0	CAF	2016	9.533511e+10	28842484	80.778969	3305.371054
1230	1882218	92	Antigua and Barbuda	ATG	371	409	10	CONCACAF	2016	1.460145e+09	94527	2.060043	15446.853319
...	...	...	...	...	...	...	...	...	...	...	...	...	...
40766	44005	73	Uzbekistan	UZB	489	470	1	AFC	2016	6.722034e+10	31847900	1179.956705	2110.667754
41035	43982	193	Vanuatu	VAN	47	47	0	OFC	2016	7.735029e+08	278330	104.312191	2779.085603
41303	43841	146	Vietnam	VIE	209	200	1	AFC	2016	2.052762e+11	93638724	7315.612793	2192.214539
41572	43891	67	Zambia	ZAM	516	481	6	CAF	2016	2.106399e+10	16363507	3.878827	1287.253990
41841	43892	108	Zimbabwe	ZIM	322	318	2	CAF	2016	1.661996e+10	14030390	0.521862	1184.568669

160 rows × 13 columns

MACRO INDICATORS vs WORLD

To get a general idea as to if there is a correlation between the economic factors we chose and the FIFA standings of the countries of the world, we will be doing a linear regression between each country's GDP, GDP-per-capita, and PPP with their total FIFA points to examine the resulting p-value. If the p-value is <=.05, there is a signficant corrrelation, and if not then there is little correlation. The resulting line and scatter plot will also be plotted to display these results visually.

#GDP vs points world 

markers = ['o', 's','p', 'P', 'x','X', 'v', '^', 'D', 'd', '1','2','3','4','8', '>', '<', 'h', 
          'H',4,5,6,7,8,9,10,11, '$A$', '$B$', '$C$', '$D$', '$E$', '$F$', '$G$', '$E$']

confederations = master_df['confederation'].drop_duplicates()


#Create the plot and invert the x-axis so lesser value rankings are seen as
#more preferable
fig,ax = plt.subplots(figsize=(16,12), dpi= 100)
#ax.invert_xaxis()

year_df=master_df

country_groups = year_df.groupby('Country Name')
#The names if the columns we will be working with
x_label = 'total_points'
y_label = 'GDP'
#for each team, plot the salary on y axis and year on x axis
for name, group in country_groups:
    sm.OLS(group[x_label], group[y_label])
    #plot year on x axis, total payroll on y axis, choose a new marker from list of markers, set label as team
    plt.scatter(group[x_label], group[y_label], label = name)
    #set notation as scientific, use math text
    plt.ticklabel_format(axis='y', style='sci', scilimits=(-3, 3), useMathText = True)


#Set the independent variable x to be the ranks of the countries in this 
#confederation and dependent variable y to be GDPs of the countries in this 
#confederation
x = year_df[x_label]
y = year_df[y_label]


reg = sm.OLS(y, sm.add_constant(x)).fit()
print('P value: ' + str(reg.pvalues[1]))
#Calculate the linear regression of the variables using numpy, set the line's
#equation, and plot it in black
lin_reg = np.polyfit(x,y,1)
slope = lin_reg[0]
y_intercept = lin_reg[1]
reg_line = slope*x + y_intercept
plt.plot(x, reg_line, color="black")

plt.title('GDP vs points World')
plt.xlabel(x_label)
plt.ylabel(y_label)
plt.legend(bbox_to_anchor=(1.0, 1.0))
plt.show()

P value: 0.008943337542328962

#GDP capita vs points world 

markers = ['o', 's','p', 'P', 'x','X', 'v', '^', 'D', 'd', '1','2','3','4','8', '>', '<', 'h', 
          'H',4,5,6,7,8,9,10,11, '$A$', '$B$', '$C$', '$D$', '$E$', '$F$', '$G$', '$E$']

confederations = master_df['confederation'].drop_duplicates()


#Create the plot and invert the x-axis so lesser value rankings are seen as
#more preferable
fig,ax = plt.subplots(figsize=(16,12), dpi= 100)
#ax.invert_xaxis()

year_df=master_df

country_groups = year_df.groupby('Country Name')

x_label = 'total_points'
y_label = 'GDP_capita'
#for each team, plot the salary on y axis and year on x axis
for name, group in country_groups:
    sm.OLS(group[x_label], group[y_label])
    #plot year on x axis, total payroll on y axis, choose a new marker from list of markers, set label as team
    plt.scatter(group[x_label], group[y_label], label = name)
    #set notation as scientific, use math text
    plt.ticklabel_format(axis='y', style='sci', scilimits=(-3, 3), useMathText = True)
  
#for k, v in year_df.iterrows():
  #plt.annotate(v['Country Name'], (v[x_label], v[y_label]))
  

#Set the independent variable x to be the ranks of the countries in this 
#confederation and dependent variable y to be GDPs of the countries in this 
#confederation
x = year_df[x_label]
y = year_df[y_label]


reg = sm.OLS(y, sm.add_constant(x)).fit()
print('P value: ' + str(reg.pvalues[1]))
#Calculate the linear regression of the variables using numpy, set the line's
#equation, and plot it in black
lin_reg = np.polyfit(x,y,1)
slope = lin_reg[0]
y_intercept = lin_reg[1]
reg_line = slope*x + y_intercept
plt.plot(x, reg_line, color="black")

plt.title('GDP per capita vs points World')
plt.xlabel(x_label)
plt.ylabel(y_label)
plt.legend(bbox_to_anchor=(1.0, 1.0))
plt.show()

P value: 9.97895565411009e-05

#PPP vs points world 

markers = ['o', 's','p', 'P', 'x','X', 'v', '^', 'D', 'd', '1','2','3','4','8', '>', '<', 'h', 
          'H',4,5,6,7,8,9,10,11, '$A$', '$B$', '$C$', '$D$', '$E$', '$F$', '$G$', '$E$']

confederations = master_df['confederation'].drop_duplicates()


#Create the plot and invert the x-axis so lesser value rankings are seen as
#more preferable
fig,ax = plt.subplots(figsize=(16,12), dpi= 100)
#ax.invert_xaxis()

year_df=master_df

country_groups = year_df.groupby('Country Name')

x_label = 'total_points'
y_label = 'PPP'
#for each team, plot the salary on y axis and year on x axis
for name, group in country_groups:
    sm.OLS(group[x_label], group[y_label])
    #plot year on x axis, total payroll on y axis, choose a new marker from list of markers, set label as team
    plt.scatter(group[x_label], group[y_label], label = name)
    #set notation as scientific, use math text
    plt.ticklabel_format(axis='y', style='sci', scilimits=(-3, 3), useMathText = True)
  
#for k, v in year_df.iterrows():
  #plt.annotate(v['Country Name'], (v[x_label], v[y_label]))
  

#Set the independent variable x to be the ranks of the countries in this 
#confederation and dependent variable y to be GDPs of the countries in this 
#confederation
x = year_df[x_label]
y = year_df[y_label]


reg = sm.OLS(y, sm.add_constant(x)).fit()
print('P value: ' + str(reg.pvalues[1]))
#Calculate the linear regression of the variables using numpy, set the line's
#equation, and plot it in black
lin_reg = np.polyfit(x,y,1)
slope = lin_reg[0]
y_intercept = lin_reg[1]
reg_line = slope*x + y_intercept
plt.plot(x, reg_line, color="black")

plt.title('PPP vs points World')
plt.xlabel(x_label)
plt.ylabel(y_label)
plt.legend(bbox_to_anchor=(1.0, 1.0))
plt.show()

P value: 0.11707892380039306

According to the graphs and their associated p values, GDP and GDP per capita have a strong correlation with points for the entire world, and PPP has an insignificant correlation. However, individual countries are hard to see at this scale, so zooming in and seperating based on confederation will allow for better a visualization and analysis.

MACRO INDICATORS vs CONFEDERATIONS

To get a better idea as to if there is a correlation between the GDP, GDP-per-capita, and PPP of a country and it's FIFA standing, we will examine the same linear regressions as before, but specific to each confederation. This way, the regional factors/circumstances affecting each country

If the p-value is <=.05, there is a signficant corrrelation, and if not then there is little correlation. The resulting line and scatter plot will also be plotted to display these results visually.

markers = ['o', 's','p', 'P', 'x','X', 'v', '^', 'D', 'd', '1','2','3','4','8', '>', '<', 'h', 
          'H',4,5,6,7,8,9,10,11, '$A$', '$B$', '$C$', '$D$', '$E$', '$F$', '$G$', '$E$']

confederations = master_df['confederation'].drop_duplicates()

for confederation in confederations:

  #Create the plot and invert the x-axis so lesser value rankings are seen as
  #more preferable
  fig,ax = plt.subplots(figsize=(16,12), dpi= 100)
  #ax.invert_xaxis()

  year_df=master_df
  year_df = master_df.loc[master_df['confederation']==confederation]

  country_groups = year_df.groupby('Country Name')

  x_label = 'total_points'
  y_label = 'GDP'
  #for each team, plot the salary on y axis and year on x axis
  for name, group in country_groups:
      sm.OLS(group[x_label], group[y_label])
      #plot year on x axis, total payroll on y axis, choose a new marker from list of markers, set label as team
      plt.scatter(group[x_label], group[y_label], label = name)
      #set notation as scientific, use math text
      plt.ticklabel_format(axis='y', style='sci', scilimits=(-3, 3), useMathText = True)
    
  for k, v in year_df.iterrows():
    plt.annotate(v['Country Name'], (v[x_label], v[y_label]))
    

  #Set the independent variable x to be the ranks of the countries in this 
  #confederation and dependent variable y to be GDPs of the countries in this 
  #confederation
  x = year_df[x_label]
  y = year_df[y_label]


  reg = sm.OLS(y, sm.add_constant(x)).fit()
  print('P value: ' + str(reg.pvalues[1]))
  #Calculate the linear regression of the variables using numpy, set the line's
  #equation, and plot it in black
  lin_reg = np.polyfit(x,y,1)
  slope = lin_reg[0]
  y_intercept = lin_reg[1]
  reg_line = slope*x + y_intercept
  plt.plot(x, reg_line, color="black")

  plt.title('GDP vs points for ' + confederation)
  plt.xlabel(x_label)
  plt.ylabel(y_label)
  plt.legend(bbox_to_anchor=(1.0, 1.0))
  plt.show()

P value: 0.018478956876983768

P value: 0.0002412840604305903

P value: 0.09060187484423693

P value: 0.05397110201669657

P value: 0.20290863689204547

P value: 0.04930348876730901

The scatter plots above plot GDP vs points for all countries, with plots seperated by their confederation. According to the p values, there is a significant correlation between GDP and points for the AFC and UEFA confederations, with all other confederations having an insignificant correlation.

markers = ['o', 's','p', 'P', 'x','X', 'v', '^', 'D', 'd', '1','2','3','4','8', '>', '<', 'h', 
           'H',4,5,6,7,8,9,10,11, '$A$', '$B$', '$C$', '$D$', '$E$', '$F$', '$G$', '$E$']

confederations = master_df['confederation'].drop_duplicates()

for confederation in confederations:
  #Create the plot and invert the x-axis so lesser value rankings are seen as
  #more preferable
  fig,ax = plt.subplots(figsize=(16,12), dpi= 100)
  #ax.invert_xaxis()

  year_df = master_df.loc[master_df['confederation']==confederation]
  country_groups = year_df.groupby('Country Name')

  x_label = 'total_points'
  y_label = 'GDP_capita'
  #for each team, plot the salary on y axis and year on x axis
  for name, group in country_groups:
      sm.OLS(group[x_label], group[y_label])
      #plot year on x axis, total payroll on y axis, choose a new marker from list of markers, set label as team
      plt.scatter(group[x_label], group[y_label], label = name)
      #set notation as scientific, use math text
      plt.ticklabel_format(axis='y', style='sci', scilimits=(-3, 3), useMathText = True)
    
  for k, v in year_df.iterrows():
    plt.annotate(v['Country Name'], (v[x_label], v[y_label]))
    

  #Set the independent variable x to be the ranks of the countries in this 
  #confederation and dependent variable y to be GDPs of the countries in this 
  #confederation
  x = year_df[x_label]
  y = year_df[y_label]


  reg = sm.OLS(y, sm.add_constant(x)).fit()
  print('P value: ' + str(reg.pvalues[1]))
  #Calculate the linear regression of the variables using numpy, set the line's
  #equation, and plot it in black
  lin_reg = np.polyfit(x,y,1)
  slope = lin_reg[0]
  y_intercept = lin_reg[1]
  reg_line = slope*x + y_intercept
  plt.plot(x, reg_line, color="black")

  plt.title('GDP per capita vs points for ' + confederation)
  plt.xlabel(x_label)
  plt.ylabel(y_label)
  plt.legend(bbox_to_anchor=(1.0, 1.0))
  plt.show()

P value: 0.0012027344565364061

P value: 0.22210985633475808

P value: 0.48349261674087574

P value: 0.10633954172818269

P value: 0.03793824215233343

P value: 0.02866526239773614

The scatter plots above plot GDP per capita vs points for all countries, with plots seperated by their confederation. According to the p values, there is a significant correlation between GDP per capita and points for the AFC and CONMEBOL confederations, with all other confederations having an insignificant correlation.

markers = ['o', 's','p', 'P', 'x','X', 'v', '^', 'D', 'd', '1','2','3','4','8', '>', '<', 'h', 
           'H',4,5,6,7,8,9,10,11, '$A$', '$B$', '$C$', '$D$', '$E$', '$F$', '$G$', '$E$']

confederations = master_df['confederation'].drop_duplicates()

for confederation in confederations:
  #Create the plot and invert the x-axis so lesser value rankings are seen as
  #more preferable
  fig,ax = plt.subplots(figsize=(16,12), dpi= 100)
  #ax.invert_xaxis()
  ax.invert_yaxis()

  year_df = master_df.loc[master_df['confederation']==confederation]
  country_groups = year_df.groupby('Country Name')

  x_label = 'total_points'
  y_label = 'PPP'
  #for each team, plot the salary on y axis and year on x axis
  for name, group in country_groups:
      sm.OLS(group[x_label], group[y_label])
      #plot year on x axis, total payroll on y axis, choose a new marker from list of markers, set label as team
      plt.scatter(group[x_label], group[y_label], label = name)
      #set notation as scientific, use math text
      plt.ticklabel_format(axis='y', style='sci', scilimits=(-3, 3), useMathText = True)
    
  for k, v in year_df.iterrows():
    plt.annotate(v['Country Name'], (v[x_label], v[y_label]))
    

  #Set the independent variable x to be the ranks of the countries in this 
  #confederation and dependent variable y to be GDPs of the countries in this 
  #confederation
  x = year_df[x_label]
  y = year_df[y_label]


  reg = sm.OLS(y, sm.add_constant(x)).fit()
  print('P value: ' + str(reg.pvalues[1]))
  #Calculate the linear regression of the variables using numpy, set the line's
  #equation, and plot it in black
  lin_reg = np.polyfit(x,y,1)
  slope = lin_reg[0]
  y_intercept = lin_reg[1]
  reg_line = slope*x + y_intercept
  plt.plot(x, reg_line, color="black")

  plt.title('PPP vs points for ' + confederation)
  plt.xlabel(x_label)
  plt.ylabel(y_label)
  plt.legend(bbox_to_anchor=(1.0, 1.0))
  plt.show()

P value: 0.48650411795804793

P value: 0.5703239890238323

P value: 0.10213476566138502

P value: 0.14676968613931507

P value: 0.5544214217577045

P value: 0.7696488294407736

The scatter plots above plot PPP vs points for all countries, with plots seperated by their confederation. According to the p values, there is a insignificant correlation between PPP and points for all confederations

To synthesize the above findings; GDP and GDP per capita had strong correlations with points for AFC

GDP had strong correlations with points for UEFA

Only GDP per capita had a strong correlation with points for CONMEBOL

PPP was insignificant for all confederations

External Factors?

Before plotting the individual confederations, the GDP and GDP-per-capita plot of all the countries might have led us to believe countries with higher values in both tend to have better performing national teams. However, now that we zoom in to the individual confederations, we see OFC, CONMEBOL, CONCACAF, and CAF do not have a low enough p-value to conclusively correlate with GDP and neither do OFC, CONCACAF, CAF, and UEFA correlate conclusively enough with GDP-per-capita. Thus, there must be another more signifigant factor at play we can search for.

To this end, we will examine the stability of each country. Our hypothesis is that, the more stable the livelihoods of the citizens of a country are, the more time some citizens might have to enjoy and develop in recreational activities like soccer. The best metric to measure this stability would be each country's fragility index.

Fragility Index

Why fragility index?

A country's Fragility Index is a measure aimed to assess their vulnerability to conflict or collapse based off certain factors (Brain drain, External intervention, etc). We chose this as a metric for a country's stability since, the higher the score, the less stable a country will be. A country's fragility index is calculated by a conflict assessment framework called CAST (Conflict Assessment System Tool) and detailed in the documentation in the following source: https://fragilestatesindex.org/indicators/

Download the Fragility index excel file from here, open in any spreadsheet program, and click save as csv with the filename 'country_stability_2016.csv': https://fragilestatesindex.org/excel/

#Read in the country stability index data from the file in this directory, drop
#all the irrelevant columns and rows
fragility_df_root = pd.read_csv("country_stability_2016.csv")
fragility_df = fragility_df_root.drop(columns=['C1: Security Apparatus',"C2: Factionalized Elites", "C3: Group Grievance", "E1: Economy", "E2: Economic Inequality",
                                          "E3: Human Flight and Brain Drain", "P1: State Legitimacy", "P2: Public Services", "P3: Human Rights", "S1: Demographic Pressures",
                                          "S2: Refugees and IDPs", "X1: External Intervention"])
fragility_df = fragility_df.drop(labels=range(178,len(fragility_df)), axis=0)
#Change the year column types to be integers and replace the Total column name
#to be "Stability_Index. Then, print the head"
fragility_df['Year'] = fragility_df['Year'].astype(int)
fragility_df.rename(columns={'Total':'fragility'}, inplace=True)
fragility_df.rename(columns={'Country':'Country Name'}, inplace=True)
fragility_df.head()

	Country Name	Year	Rank	fragility
0	Somalia	2016	1st	114.0
1	South Sudan	2016	2nd	113.8
2	Central African Republic	2016	3rd	112.1
3	Sudan	2016	4th	111.5
4	Yemen	2016	4th	111.5

master_df = master_df.merge(fragility_df, how='inner', on=['Country Name', 'Year'])
master_df.head()

	id	rank	Country Name	country_abrv	total_points	previous_points	rank_change	confederation	Year	GDP	Population	PPP	GDP_capita	Rank	fragility
0	20000	153	Afghanistan	AFG	183	194	3	AFC	2016	1.946902e+10	35383128	17.445828	550.234626	9th	107.9
1	43932	38	Albania	ALB	746	746	0	UEFA	2016	1.186387e+10	2876101	42.385879	4124.982390	124th	61.2
2	43843	28	Algeria	ALG	845	845	0	CAF	2016	1.590491e+11	40551404	37.155827	3922.160050	76th	78.3
3	43844	105	Angola	ANG	333	330	0	CAF	2016	9.533511e+10	28842484	80.778969	3305.371054	37th	90.5
4	1882218	92	Antigua and Barbuda	ATG	371	409	10	CONCACAF	2016	1.460145e+09	94527	2.060043	15446.853319	129th	56.2

Make a scatter plot of each confederation's fragility score as opposed to their total FIFA points. Invert the fragility score axis so that lesser fragility indicies (higher stability) are shown as possitive ascension and plot a linear regression to show how the data trends. Likewise, print the resulting p-values of the regression for each plot to show if the countries' fragility index correlates with their total FIFA points within that confederation.

markers = ['o', 's','p', 'P', 'x','X', 'v', '^', 'D', 'd', '1','2','3','4','8', '>', '<', 'h', 
           'H',4,5,6,7,8,9,10,11, '$A$', '$B$', '$C$', '$D$', '$E$', '$F$', '$G$', '$E$']

confederations = master_df['confederation'].drop_duplicates()
fragility_pvalues = []
for confederation in confederations:

    #Create the plot and invert the x-axis so lesser value rankings are seen as
    #more preferable
  fig,ax = plt.subplots(figsize=(16,12), dpi= 100)
  # ax.invert_xaxis()
  ax.invert_yaxis()

  #year_df=master_df
  year_df = master_df.loc[master_df['confederation']==confederation]

  country_groups = year_df.groupby('Country Name')

  x_label = 'total_points'
  y_label = 'fragility'
  #for each team, plot the salary on y axis and year on x axis
  for name, group in country_groups:
      sm.OLS(group[x_label], group[y_label])
      #plot year on x axis, total payroll on y axis, choose a new marker from list of markers, set label as team
      plt.scatter(group[x_label], group[y_label], label = name)
      #set notation as scientific, use math text
      plt.ticklabel_format(axis='y', style='sci', scilimits=(-3, 3), useMathText = True)
    
  for k, v in year_df.iterrows():
    plt.annotate(v['Country Name'], (v[x_label], v[y_label]))
    

  #Set the independent variable x to be the ranks of the countries in this 
  #confederation and dependent variable y to be GDPs of the countries in this 
  #confederation
  x = year_df[x_label]
  y = year_df[y_label]


  reg = sm.OLS(y, sm.add_constant(x)).fit()
  print('P value: ' + str(reg.pvalues[1]))

  fragility_pvalues.append(confederation + ': ' + str(reg.pvalues))

  #Calculate the linear regression of the variables using numpy, set the line's
  #equation, and plot it in black
  lin_reg = np.polyfit(x,y,1)
  slope = lin_reg[0]
  y_intercept = lin_reg[1]
  reg_line = slope*x + y_intercept
  plt.plot(x, reg_line, color="black")

  plt.title('Fragility Index vs points for ' + confederation)
  plt.xlabel(x_label)
  plt.ylabel(y_label)
  plt.legend(bbox_to_anchor=(1.0, 1.0))
  plt.show()

P value: 0.004762090342674282

P value: 0.02057164340325089

P value: 0.7584063816021708

P value: 0.3138278202314517

P value: 0.1380979183559405

P value: 0.022709373026780114

The scatter plots above plot fragility index vs points for all countries, with plots seperated by their confederation. According to the p values, there is a significant correlation between the fragility index and points for the AFC, UEFA, CAF, CONCACAF, and OFC confederations, with only the CONMEBOL confederation having an insignificant correlation.

heatmap_df=master_df[['total_points', 'GDP', 'Population', 'PPP', 'Rank', 'fragility',
       'GDP_capita']]
heatmap_df

	total_points	GDP	Population	PPP	Rank	fragility	GDP_capita
0	183	1.946902e+10	35383128	17.445828	9th	107.9	550.234626
1	746	1.186387e+10	2876101	42.385879	124th	61.2	4124.982390
2	845	1.590491e+11	40551404	37.155827	76th	78.3	3922.160050
3	333	9.533511e+10	28842484	80.778969	37th	90.5	3305.371054
4	371	1.460145e+09	94527	2.060043	129th	56.2	15446.853319
...	...	...	...	...	...	...	...
143	1074	5.241972e+10	3424132	22.453821	155th	36.2	15308.907692
144	489	6.722034e+10	31847900	1179.956705	60th	83.5	2110.667754
145	209	2.052762e+11	93638724	7315.612793	106th	70.7	2192.214539
146	516	2.106399e+10	16363507	3.878827	49th	86.3	1287.253990
147	322	1.661996e+10	14030390	0.521862	16th	100.5	1184.568669

148 rows × 7 columns

fig,ax = plt.subplots(figsize=(12,8), dpi= 100)

corr=heatmap_df.corr()
sns.heatmap(corr, annot=True)
plt.show

<function matplotlib.pyplot.show>

The above heatmap visualizes the correelations between total poiints and fragility for all teams, with strong correlations having dark or light colorations, depending on the direction of the correlation. This verifies our observation made previous, that the fragility index has the most signficant correlation to total points of a soccer team, because fragility has the highest absolute value of all the other indicators when comparing with total points.

Our initial hypothesis has been almost completley rebuked, as GPD, GDP per capita, and PPP have proven to be largely insignificant. However, fragility index appears to have some signficance, so it may be worthwhile to investigate this aspect further. The fragility index is an index, meaning it is comprised of multiple other micro-indicators, so by comparing these micro-indicators with total points, we could isolate the specific indicators that yield the best correlation with total points.

fragility_heatmap_df = fragility_df_root[['Country','Total', 'C1: Security Apparatus',"C2: Factionalized Elites", "C3: Group Grievance", "E1: Economy", "E2: Economic Inequality",
                                          "E3: Human Flight and Brain Drain", "P1: State Legitimacy", "P2: Public Services", "P3: Human Rights", "S1: Demographic Pressures",
                                          "S2: Refugees and IDPs", "X1: External Intervention"]]

fragility_heatmap_df.rename(columns={'Country':'Country Name'}, inplace=True)
fragility_heatmap_df = fragility_heatmap_df.merge(master_df[['Country Name','total_points']], how='inner', on=['Country Name'])
fragility_heatmap_df
                               

/usr/local/lib/python3.7/dist-packages/pandas/core/frame.py:4308: SettingWithCopyWarning: 
A value is trying to be set on a copy of a slice from a DataFrame

See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy
  errors=errors,

	Country Name	Total	C1: Security Apparatus	C2: Factionalized Elites	C3: Group Grievance	E1: Economy	E2: Economic Inequality	E3: Human Flight and Brain Drain	P1: State Legitimacy	P2: Public Services	P3: Human Rights	S1: Demographic Pressures	S2: Refugees and IDPs	X1: External Intervention	total_points
0	Somalia	114.0	9.7	10.0	9.4	9.0	9.3	9.5	9.5	9.0	9.7	9.7	9.7	9.5	0
1	Central African Republic	112.1	9.2	10.0	9.3	8.6	9.9	7.2	9.8	10.0	9.9	8.7	10.0	9.5	302
2	Sudan	111.5	9.2	10.0	9.8	8.7	7.6	9.1	9.8	9.1	9.3	9.0	10.0	9.9	271
3	Chad	110.1	9.1	9.8	8.5	8.0	9.3	8.9	9.2	9.8	9.3	9.9	9.8	8.5	301
4	Afghanistan	107.9	10.0	8.6	8.6	8.5	7.5	8.4	9.1	9.6	8.7	9.5	9.5	9.9	183
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
143	Switzerland	21.8	1.3	1.0	3.8	2.5	2.4	1.8	1.0	1.1	2.0	1.7	2.2	1.0	1050
144	Denmark	21.5	1.4	1.4	4.6	2.2	1.8	1.6	0.6	1.1	1.7	2.2	1.8	1.1	685
145	New Zealand	21.3	1.4	1.1	3.8	3.8	2.5	1.8	0.6	1.6	0.7	1.6	1.5	0.9	192
146	Norway	21.2	2.2	1.1	3.8	1.7	1.7	1.3	0.5	1.1	1.6	1.7	3.0	1.5	591
147	Finland	18.8	1.4	1.1	2.0	3.7	1.2	2.0	0.6	1.0	1.0	1.3	2.5	1.0	663

148 rows × 15 columns

To quickly visualize the correlations between all of the micro-indicators of the Fragility index with total points, we will utilize a heatmap.

fig,ax = plt.subplots(figsize=(12,8), dpi= 100)

corr=fragility_heatmap_df.corr()
sns.heatmap(corr, annot=True)
plt.show

<function matplotlib.pyplot.show>

for i in master_df.columns:
  if i == 'fragility':
    print('\n'+i)
    x = master_df['total_points']
    y = master_df[i]
    reg = sm.OLS(y, sm.add_constant(x)).fit()
    print(reg.pvalues)

fragility
const           2.348571e-59
total_points    1.247753e-08
dtype: float64

This heatmap shows the 2 indicators that best correlate with total FIFA ranking points is X1: External Intervention and E3: Human flight and Brain Drain. Both have a coefficient of -0.48, which when calculating as absolute value is .03 higher than the baseline coefficient of -0.45. Therefore, these 2 indicators should result in a better correlation with total points than just using fragility index.

for i in fragility_heatmap_df.columns:
  if i != 'Country Name':
    print('\n'+i)
    x = fragility_heatmap_df['total_points']
    y = fragility_heatmap_df[i]

    reg = sm.OLS(y, sm.add_constant(x)).fit()
    print(reg.pvalues)

Total
const           2.348571e-59
total_points    1.247753e-08
dtype: float64

C1: Security Apparatus
const           2.252684e-48
total_points    1.330899e-06
dtype: float64

C2: Factionalized Elites
const           7.890897e-53
total_points    2.509815e-07
dtype: float64

C3: Group Grievance
const           4.166836e-56
total_points    1.131515e-03
dtype: float64

E1: Economy
const           1.795711e-56
total_points    5.831602e-05
dtype: float64

E2: Economic Inequality
const           4.708398e-54
total_points    5.615439e-06
dtype: float64

E3: Human Flight and Brain Drain
const           2.973784e-54
total_points    5.940894e-10
dtype: float64

P1: State Legitimacy
const           5.382668e-49
total_points    2.038916e-07
dtype: float64

P2: Public Services
const           5.717766e-44
total_points    5.367091e-07
dtype: float64

P3: Human Rights
const           2.966822e-50
total_points    3.592877e-08
dtype: float64

S1: Demographic Pressures
const           7.569971e-50
total_points    1.277882e-07
dtype: float64

S2: Refugees and IDPs
const           5.464047e-38
total_points    1.644265e-02
dtype: float64

X1: External Intervention
const           5.064455e-52
total_points    4.498906e-10
dtype: float64

total_points
const           1.274721e-09
total_points    0.000000e+00
dtype: float64

The p values above all confirm what we found in the heatmap; X1: External Intervention and E3: Human flight and Brain Drain have the strongest correlation with total points. However, E3: Human flight and Brain Drain has a lower p-value and therefore stronger correlation than X1: External Intervention, therefore E3 should be the canidate for further examination to see if it can be a better indicator than Fragility index itself.

BRAIN DRAIN

"E3: Human flight and Brain Drain" is defined as the emigration of the economically productive populations of a country due to economic deterioration in their home country and the hope of better opportunities afar. Merge this information into the table using the same function as before and display the head of the table to show the change.

#Add E3: Human flight and Brain Drain to master_df as E3 Brain drain
master_df = master_df.merge(fragility_heatmap_df[['E3: Human Flight and Brain Drain', 'Country Name']], how='inner', on=['Country Name'])
master_df = master_df.rename(columns={"E3: Human Flight and Brain Drain": "E3 Brain Drain"})
master_df = master_df.loc[:,~master_df.columns.duplicated()]

master_df.head()

	id	rank	Country Name	country_abrv	total_points	previous_points	rank_change	confederation	Year	GDP	Population	PPP	GDP_capita	Rank	fragility	E3 Brain Drain
0	20000	153	Afghanistan	AFG	183	194	3	AFC	2016	1.946902e+10	35383128	17.445828	550.234626	9th	107.9	8.4
1	43932	38	Albania	ALB	746	746	0	UEFA	2016	1.186387e+10	2876101	42.385879	4124.982390	124th	61.2	7.6
2	43843	28	Algeria	ALG	845	845	0	CAF	2016	1.590491e+11	40551404	37.155827	3922.160050	76th	78.3	5.4
3	43844	105	Angola	ANG	333	330	0	CAF	2016	9.533511e+10	28842484	80.778969	3305.371054	37th	90.5	6.3
4	1882218	92	Antigua and Barbuda	ATG	371	409	10	CONCACAF	2016	1.460145e+09	94527	2.060043	15446.853319	129th	56.2	7.0

We will analyze the Brain Drain indicator as we have analyzed our previous indicators, with a linear regression seperated for all confederations, so our analysis is methodically consistent and we are not introducing any new regressional skews.

markers = ['o', 's','p', 'P', 'x','X', 'v', '^', 'D', 'd', '1','2','3','4','8', '>', '<', 'h', 
           'H',4,5,6,7,8,9,10,11, '$A$', '$B$', '$C$', '$D$', '$E$', '$F$', '$G$', '$E$']

confederations = master_df['confederation'].drop_duplicates()

e3_pvalues = []
for confederation in confederations:

    #Create the plot and invert the x-axis so lesser value rankings are seen as
    #more preferable
  fig,ax = plt.subplots(figsize=(16,12), dpi= 100)
  # ax.invert_xaxis()
  ax.invert_yaxis()

  #year_df=master_df
  year_df = master_df.loc[master_df['confederation']==confederation]

  country_groups = year_df.groupby('Country Name')

  x_label = 'total_points'
  y_label = 'E3 Brain Drain'
  #for each team, plot the salary on y axis and year on x axis
  for name, group in country_groups:
      sm.OLS(group[x_label], group[y_label])
      #plot year on x axis, total payroll on y axis, choose a new marker from list of markers, set label as team
      plt.scatter(group[x_label], group[y_label], label = name)
      #set notation as scientific, use math text
      plt.ticklabel_format(axis='y', style='sci', scilimits=(-3, 3), useMathText = True)
    
  for k, v in year_df.iterrows():
    plt.annotate(v['Country Name'], (v[x_label], v[y_label]))
    

  #Set the independent variable x to be the ranks of the countries in this 
  #confederation and dependent variable y to be GDPs of the countries in this 
  #confederation
  x = year_df[x_label]
  y = year_df[y_label]


  reg = sm.OLS(y, sm.add_constant(x)).fit()
  print('P values, consider rank only \n' + str(reg.pvalues))
  e3_pvalues.append(confederation + ': ' + str(reg.pvalues))
  #Calculate the linear regression of the variables using numpy, set the line's
  #equation, and plot it in black
  lin_reg = np.polyfit(x,y,1)
  slope = lin_reg[0]
  y_intercept = lin_reg[1]
  reg_line = slope*x + y_intercept
  plt.plot(x, reg_line, color="black")

  plt.title('Brain Drain vs points for ' + confederation)
  plt.xlabel(x_label)
  plt.ylabel(y_label)
  plt.legend(bbox_to_anchor=(1.0, 1.0))
  plt.show()

P values, consider rank only 
const           9.640533e-15
total_points    4.500124e-04
dtype: float64

P values, consider rank only 
const           1.264805e-11
total_points    3.015223e-03
dtype: float64

P values, consider rank only 
const           1.041262e-19
total_points    5.770920e-01
dtype: float64

P values, consider rank only 
const           1.211796e-07
total_points    4.009242e-02
dtype: float64

P values, consider rank only 
const           0.000135
total_points    0.014485
dtype: float64

P values, consider rank only 
const           0.022346
total_points    0.325008
dtype: float64

According to the above graphs and their p-values, Brain Drain the greatest signficance to the countries of all confederations except OFC. And since OFC consists of fewer representative nations as CONMEBOL, it would be considered a better overall indicator that the correlation increases for all other confederations despite the drawback in this confederation.

cnt=0
for i in e3_pvalues:
  print('Brain Drain p value')
  print(i)
  print('Fragility Index p value')
  print(fragility_pvalues[cnt])
  print()
  cnt+=1

Brain Drain p value
AFC: const           9.640533e-15
total_points    4.500124e-04
dtype: float64
Fragility Index p value
AFC: const           2.067913e-16
total_points    4.762090e-03
dtype: float64

Brain Drain p value
UEFA: const           1.264805e-11
total_points    3.015223e-03
dtype: float64
Fragility Index p value
UEFA: const           2.315754e-11
total_points    2.057164e-02
dtype: float64

Brain Drain p value
CAF: const           1.041262e-19
total_points    5.770920e-01
dtype: float64
Fragility Index p value
CAF: const           1.177884e-18
total_points    7.584064e-01
dtype: float64

Brain Drain p value
CONCACAF: const           1.211796e-07
total_points    4.009242e-02
dtype: float64
Fragility Index p value
CONCACAF: const           2.132769e-07
total_points    3.138278e-01
dtype: float64

Brain Drain p value
CONMEBOL: const           0.000135
total_points    0.014485
dtype: float64
Fragility Index p value
CONMEBOL: const           0.001086
total_points    0.138098
dtype: float64

Brain Drain p value
OFC: const           0.022346
total_points    0.325008
dtype: float64
Fragility Index p value
OFC: const           0.001356
total_points    0.022709
dtype: float64

Comparing our p values shows that Brain Drain has increased accuracy for the UEFA, CONCACAF, and CONMEBOL, and has decreased accuracy for AFC and OFC. The decrease in accuracy for AFC and increase for UEFA, CAF, and CONCACAF is negligable, because all of their p values are very close to 0 regardless. CONMEBOL has become signficantly more accurate, as the p value has went from 0.138098, an insignificant value, to 0.014485, a signficant value. Contrastingly, the p values for OFC went from 0.022709 to 0.325008, losing its significance.

Conclusions

Takeaways:

• PPP (Purchasing Power Parity) is not a signifigant indicator of a country's national team performances at any level (global or regional).
• The GDP (Gross Domestic Product) and GDP per capita of a country are possitively correlated with their men's national team performances at a global level, but inconsistent at a regional level.
• The fragility index of a country is a signifigant indicator of a negative correlation between a country's instability and their men's national team's performance (and conversely, a possitive correlation between a country's stability and performance) for all confederations other than CONMEBOL.
• The Brain Drain of a country is comparatively the most signifigant indicator of a negative correlation between itself and the performance of that country's national team across all confederations aside from the outlier OFC confederation.

Future Possibilities

The analytic method we used to quantify the signficance of an indicator was rather simple; Isolate the variable to its own column, compare the variable against the potential dependent variable via a linear regression, and check the significance with p-value. We were consistent with our methodolgy, which limits the extent of unnecassary skews due to a change in analytic approach. And even though we didnt find a definitive set of indicators that could correlate to a country's football performance, we havent analyzed everything. There are many other factors, from climate conditions to government spending breakdowns that could be attributed to football performance. Additionally, there are many other models that could yield better predictions than a linear regression, such as a polynomial regression, SVM, or even a neural network. As a final takeaway, this tutorial is a mere stepping stone to the wide range of analytical approaches one could take to explain what aspects of a country most influences its professional football performace.